基于前3行其他列值创建Y/N型标记列

Alright, let's break down how to recreate those two flag columns—for diabetes and hypertension—from your mortality dataset, focusing only on 2010-2016 data and checking the first 3 ENICON columns (ENICON1 to ENICON3) as specified.

Step 1: Filter to Target Years #

First, narrow down your dataset to only the years you care about:

import pandas as pd

# Replace 'mortality_data' with your actual DataFrame name
filtered_data = mortality_data[(mortality_data['year'] >= 2010) & (mortality_data['year'] <= 2016)]

Step 2: Define Your Target Codes #

You'll need to map the ICD codes (or whatever encoding your dataset uses) for diabetes and hypertension. I'll use common ICD-10 codes as examples—swap these out for the actual values in your ENICON columns:

# Example ICD-10 codes for diabetes
diabetes_codes = ['E10', 'E11', 'E12', 'E13', 'E14']
# Example ICD-10 codes for hypertension
hypertension_codes = ['I10', 'I11', 'I12', 'I13']

Step 3: Create the Flag Columns #

The core logic here is: check if any of the first 3 ENICON columns contain a target code, then mark the flag as 1 (yes) or 0 (no).

Diabetes Flag #

# Check ENICON1-3 for diabetes codes; return 1 if any match, 0 otherwise
filtered_data['diabetes_flag'] = filtered_data[['ENICON1', 'ENICON2', 'ENICON3']].isin(diabetes_codes).any(axis=1).astype(int)

Hypertension Flag #

# Repeat the logic for hypertension
filtered_data['hypertension_flag'] = filtered_data[['ENICON1', 'ENICON2', 'ENICON3']].isin(hypertension_codes).any(axis=1).astype(int)

Step 4: Verify Your Results #

To make sure the flags are working as expected, spot-check some rows:

# View the first 10 rows with the relevant columns
print(filtered_data[['ENICON1', 'ENICON2', 'ENICON3', 'diabetes_flag', 'hypertension_flag']].head(10))

SAS Logic Reference #

For context, here's what the equivalent SAS code would look like, matching your original logic:

data filtered_data;
    set mortality_data;
    where year between 2010 and 2016;
    /* Initialize flags to 0 (no) */
    diabetes_flag = 0;
    hypertension_flag = 0;
    /* Set diabetes flag if any of first 3 ENICON columns have a match */
    if ENICON1 in ('E10','E11','E12','E13','E14') or 
       ENICON2 in ('E10','E11','E12','E13','E14') or 
       ENICON3 in ('E10','E11','E12','E13','E14') then diabetes_flag = 1;
    /* Set hypertension flag similarly */
    if ENICON1 in ('I10','I11','I12','I13') or 
       ENICON2 in ('I10','I11','I12','I13') or 
       ENICON3 in ('I10','I11','I12','I13') then hypertension_flag = 1;
run;

内容的提问来源于stack exchange，提问作者Mariana Cendon